by Chris DePuy, 650 Group / October 16, 2026
The OCP Global Summit opened Monday in San Jose and closed on October 15, sold out for 2026, and the momentum behind the show keeps building. The organizers are reporting attendance of roughly 15,000, with about 3,000 more would-be attendees turned away because the show could not fit them — demand at a level the San Jose edition has never seen. We have been counting the companies that time press releases around each major trade show, and the OCP arc is one of the cleanest growth stories on the calendar: show-named company releases went from 9 in 2023 to 14 in 2024 to 27 last year, and attendance climbed from 4,401 to 7,047 to 10,835 over the prior two stretches, roughly two and a half times in two years, per the show organizer’s own accounting. If the organizers’ early attendance read of roughly 15,000 holds, this year’s gate count would be up about 38 percent from 2025 — with about 3,000 more turned away for space, demand is outrunning the venue. The pre-2024 readings run thin in the news archives we count from, so we treat those early numbers as floors rather than precise counts, and this year’s harvest runs picked up releases earlier passes had missed (ASUS, Credo, and LITEON in 2023, MSI in 2024, and Alcor Micro, LITEON, MiTAC Computing, Giga Computing, and a second Boyd release in 2025), so the 2025 figure is 27 rather than the 20 we published last October — the direction is not in dispute.
The show also keeps widening its AI footprint. In the 58 show-named releases we count for 2026, AI was in more than nine of every ten of them — 55 of 58 in our count — and the mix has shifted toward networking, which is running at nearly three in ten of the releases versus 22 percent for all of 2025. The flow was led by the SONiC Foundation welcoming new members to its growing AI networking ecosystem and the OIF showcasing its technical work advancing AI infrastructure, and the show-week wave brought the expected server, rack, cooling, and power announcements, led by NVIDIA opening the Vera Rubin rack-scale architecture and Cisco’s new Silicon One G302 and joint USFP pluggable-optics form factor with NVIDIA. The count below is our final tally for the 2026 summit, covering releases through the full count window into the week after the show closed. Here is the release list, in the order they landed:
- Lattice to showcase open control plane innovations for AI infrastructure — control-plane silicon aimed at AI racks (September 28)
- Murata Manufacturing showcasing ‘Powering AI Scalability’ — power components for AI build-outs (September 28)
- Compal bringing NVIDIA AI factory infrastructure to the summit — megawatt-scale power to advanced cooling and accelerated computing (September 29)
- AMD announcing CEO Lisa Su will deliver the summit’s opening keynote — on delivering the open, full-stack infrastructure required for agentic AI (September 29)
- Pacific Rubber & Packing taking VeriClean seals to the summit — sealing and containment for liquid-cooled environments (September 29)
- Lumentum announcing its speaker lineup — high-density interconnect sessions on the agenda (October 1)
- OIF showcasing technical work advancing AI infrastructure at TEF 2026 and the summit — 448G, AI optical interconnects, and a half-day AI-infrastructure workshop (October 1)
- Intel previewing its open AI infrastructure program for the summit — a keynote on open AI infrastructure plus booth demonstrations (October 5)
- Airsys and Shell showcasing spray liquid cooling for data centers (October 5)
- Ciena preparing a data center and AI interconnect showcase (October 5)
- SONiC Foundation welcoming new members to its growing AI networking ecosystem and expanding its presence at the summit (October 5)
- AEWIN unveiling flagship DC-MHS AI servers with two-phase direct liquid cooling — rack-scale cooling for racks beyond 200 kW (October 6)
- Hillcrest attending to discuss single-stage ZVS power for 800V AI data centers — modeled at about 99 percent AC-DC conversion efficiency (October 6)
- Valeo presenting its aluminum heat exchanger cooling technology with OVH Cloud — 38 percent more cooling capacity and 25 percent lower fan power versus copper (October 7)
- Nidec announcing its first-ever OCP appearance with its coolant distribution unit line — an 800VDC-compatible In-Row CDU and a 4U 300kW In-Rack CDU (October 7)
- Marvell to showcase open AI infrastructure solutions — a 102.4T Teralynx data center switch, CXL memory, and optical interconnect demos (October 8)
- Delta showcasing energy-saving grid-to-chip infrastructure for sustainable AI — power, cooling, and infrastructure at booth B5 (October 8)
- KIOXIA showcasing game-changing scalability for AI data storage — XL-FLASH Gen 3 SSDs and liquid-cooled E1.S drives for AI servers (October 8)
- Acbel to showcase high-density power solutions for AI data centers — 100kW and 72kW power shelf solutions (October 8)
- Broadcom to showcase AI networking and co-packaged optics — Tomahawk 6, Tomahawk Ultra, and Jericho 4 switches, Thor Ultra NICs, and third-generation TH6-Davisson co-packaged optics, with a keynote from Asad Khamisy (October 8)
- Credo to showcase a broad portfolio of connectivity solutions for AI scale-out and scale-up fabrics (October 9)
- Keysight showcasing AI infrastructure validation advances across scale-up, scale-out, and scale-across networks to inference (October 9)
- KYZEN’s Chelsea Jewell presenting on pre-immersion cleaning chemistries — chemistry that shapes long-term reliability of liquid-cooled AI server boards (October 9)
- Hemmersbach joining hyperscalers, OEMs, and data center operators at the summit (October 9)
- NVIDIA opening the Vera Rubin rack-scale architecture to help scale AI factories worldwide — contributed OCP-style designs as AI factories grow from megawatts to gigawatts (October 12)
- Cisco announcing Silicon One G302 1.6T silicon and, with NVIDIA, the new USFP 400G-per-lane pluggable optics form factor — scale-across and scale-out open Ethernet foundations for AI factories (October 12)
- Micas Networks unveiling three 102.4T open Ethernet platforms for AI factories — Tomahawk 6-based switches plus Tomahawk 6-Davisson 102.4T CPO (October 12)
- MiTAC Computing showcasing full-stack innovation to scale open AI infrastructure — AI-ready, liquid-cooled AMD-platform servers at booth D22 (October 12)
- QCT debuting new server platforms powered by AMD EPYC 9006 — a fully liquid-cooled 8-GPU D76W-2U for agentic AI plus an S46V-1U for cloud and HPC (October 12)
- QCT showcasing future-ready Intel-based data center solutions — QuantaGrid AI platforms with Intel Xeon 6 and Arc Pro B70 inference (October 12)
- QCT accelerating AI factory and distributed AI deployments with co-designed infrastructure — rack-scale, DC-MHS-aligned deployments co-designed with partners (October 12)
- ASUS unveiling the XA P2N-E2 powered by NVIDIA’s Vera CPU for agentic AI workloads (October 12)
- Delta highlighting its collaboration with NVIDIA on energy-efficient 800 VDC power and cooling for AI data centers — 800 VDC architecture for next-gen AI factories (October 15)
- ASRock Rack showcasing AI infrastructure powered by NVIDIA technologies — NVIDIA MGX and GB300-based systems (October 12)
- MSI bringing NVIDIA MGX rack-scale AI infrastructure to the summit — a 48RU rack with four liquid-cooled CG681 systems, Spectrum-4 switching, and a DGX Station-based desktop AI supercomputer (October 12)
- Lumentum, Astera Labs, and Corning demonstrating 1060 nm VCSEL-based PCIe over optics for AI rack-scale connectivity (October 12)
- Keysight showcasing Link Layer Retry interoperability with Alibaba Cloud — first multi-vendor LLR and CBFC demonstrations with Broadcom and Alibaba Cloud on 1.6T and 400GE (October 12)
- Larch Networks and Dorado showcasing a commercially available Ethernet AI fabric and orchestration (October 12)
- Vuzix showcasing the Causeway waveguide optical interconnect evaluation platform at the summit and the PECC Summit — 19,200 parallel optical pathways for CPO and near-package optics evaluation (October 12)
- Wiwynn showcasing next-generation AI platforms, thermal engineering, and optical interconnect technologies — among the first NVIDIA Vera Rubin NVL72 pilot build partners, AMD Helios rack integration, diamond-based microchannel cooling, and co-packaged optics at booth C74 (October 12)
- Wistron showcasing its next-generation AI compute and storage systems at the summit — Vera Rubin NVL72 readiness, the NVIDIA Vera CPU rack for agentic AI, Spectrum-X Ethernet Photonics, and the HGX Rubin NVL8 platform (October 12)
- SolarEdge showcasing MVAC-to-800 VDC AI factory power technology at the NVIDIA MGX ecosystem showcase — a solid-state transformer cell and containerized powertrain model for grid-to-rack 800 VDC (October 13)
- Industry leaders formally launching the SCMP OCP workstream for open, secure AI infrastructure management and control — a vendor-neutral architecture for secure control, boot, and management of data center infrastructure, announced in a joint summit keynote (October 13)
- MSI showcasing its open data center and AI infrastructure at the summit — new 21-inch ORv3 systems and DC-MHS modular building blocks alongside NVIDIA MGX racks at booth A49 (October 13)
- CyberTAN unveiling the Pallas series liquid-cooled 102.4 Tbps Ethernet switch platform blueprint, developed with Marvell — a fully liquid-cooled platform built around Marvell’s Teralynx T100 silicon for AI clusters (October 13)
- Wiwynn, Ayar Labs, and GUC demonstrating an integrated optical compute system at the summit — a live co-packaged optics demonstration pairing GUC’s ASIC with Ayar Labs’ TeraPHY optical engine and SuperNova light source (October 13)
- UfiSpace demonstrating 1.6T AI fabric thermal feasibility with a live liquid cooling platform — direct liquid cooling on its open, disaggregated networking platforms at booth D19 (October 13)
- Giga Computing bringing its rack-scale AI portfolio to the show floor at GIGABYTE booth A39 — a fully configured NVIDIA Vera Rubin NVL72 rack, nodes from a Vera CPU rack and a Groq 3 LPX rack, and an OCP ORv3 rack (October 12)
- LITEON showcasing its NVIDIA 800 VDC architecture and thermal management portfolio — 800 VDC power shelves integrated with liquid cooling for megawatt-scale AI infrastructure (October 13)
- Compal accelerating time-to-AI across NVIDIA platform generations — from the OG430-2-L1 HGX B300 server to the OG231-2-L1 HGX Rubin NVL8 and a 44U Rubin NVL8 rack, plus a Vera CPU tray and Spectrum-X SN6600 (October 13)
- Pegatron bringing AI infrastructure from design to deployment — the RA4803-72N3 Vera Rubin NVL72 rack, a liquid-cooled Vera CPU system, Groq 3 LPX, and AS210-2T1-8H3 HGX Rubin NVL8 systems (October 14)
- xFusion converging cooling, power, and interconnect for the agentic era — FusionPod for AI V8, a 42U liquid-cooled rack supporting up to 155 kW, alongside 800 VDC power and a keynote (October 14)
- BluSky AI’s chief AI officer speaking on open AI infrastructure standards — on the ‘AI Computing Continuum’ panel at the summit (October 14)
- XCENA showcasing CXL memory innovations for AI inference — SK hynix’s CXL Hybrid Memory built on XCENA’s MX1 chip, on display in the SK hynix booth (October 15)
- LITEON showcasing its NVIDIA MGX ecosystem role with an 800 VDC power rack and an 800V-to-50V power brick — 330kW, 880kW, and 1.2MW power deployment scenarios at its booth (October 15)
- SK hynix recapping its memory-centric server-systems showcase — a second-generation 256GB CMM-DDR5 on CXL 3.2 plus MARS and CMM-Ax system demos (October 16)
- Clivet showcasing its next-generation magnetic-bearing centrifugal compressor for AI cooling — Midea Building Technologies’ Clivet brand aimed at the AI computing surge (October 16)
- LITEON and QCT jointly showcasing rack-level power for the NVIDIA Vera Rubin NVL72 platform — LITEON’s 110kW power shelf integrated in the rack power architecture (October 16)
The show itself is changing addresses after this year: beginning in 2027, the OCP Global Summit moves to the Moscone Center in downtown San Francisco, with the 2027 edition set for October 4–7. A sold-out San Jose run and a move to a larger venue is the classic signature of a show that outgrew its room.
We think the OCP Global Summit is one of the most important venues on the calendar for AI-related announcements. The evidence is in the timing: Lumentum put its summit sessions on the wire a full eleven days before the doors opened, companies are increasingly synchronizing their product announcements with the show, the same pattern we count at OFC and ECOC, and the pre-show release flow is starting earlier and carrying more AI content each year. For anyone tracking AI infrastructure — from power and cooling through scale-up networking — the release window around this summit has become a primary source, not a sideshow.
650 Group researches several of the technologies we covered in this blog, including AI Servers and Data Center Interconnect and more.
